1 |
Michael Sullivan wrote: |
2 |
> I'm fed up with this. I've restarted the emerge of gcc-4.1.1-r3 on my |
3 |
> server box four times now. It always gets to compiling insn-attrtab.o |
4 |
> and then it never stops. I even went into the Makefile to find the |
5 |
> exact command used to compile this particular object file and ran the |
6 |
> command manually. It only took half an hour, but when I told it to |
7 |
> emerge sys-devel/gcc, I guess it didn't like my manually built file, so |
8 |
> it's doing it over again. This time it's been trying to build this file |
9 |
> for over eight hours. |
10 |
> |
11 |
> 4499 root 18 0 132m 35m 416 R 1.5 60.2 8:41.94 cc1 |
12 |
|
13 |
I once experienced something similar on my [i686 Pentium II (Klamath)] |
14 |
server. I don't remember exactly what part of gcc was compiling at the |
15 |
time of the "hang", but I found switching kernel versions helped. |
16 |
According to top, kswapd0 was very busy, basically using up most of the |
17 |
systems resources. Ssh'ing in was very slow due to that and the gcc |
18 |
compile. I stopped and restarted it twice after letting it run over |
19 |
night (8+ hrs). Normal compile time is "genlop -t gcc" about 4hrs and |
20 |
15mins. |
21 |
|
22 |
Googling kswapd0 showed it's related to the kernel swapping pages in and |
23 |
out. My system was swapping like nuts, but shouldn't have. |
24 |
This is a good example of that happening : |
25 |
http://lists.centos.org/pipermail/centos/2005-May/046804.html |
26 |
|
27 |
You might want to try rebooting, or changing swapiness values to see if |
28 |
that helps. In my case there was a gentoo-sources update at the same |
29 |
time, so I just let that build and restarted. |
30 |
|
31 |
> 4499 root 18 0 132m 35m 416 R 1.5 60.2 8:41.94 cc1 |
32 |
|
33 |
According to this cc1 is only using 1.5% of cpu, and about 60% of ram. |
34 |
What does "free" give you, and does kswapd0 keep jumping up and down in |
35 |
terms of cpu%? |
36 |
|
37 |
|
38 |
PaulNM |
39 |
-- |
40 |
gentoo-user@g.o mailing list |